Privacy Policy

Last updated: July 30, 2026

Tokscale is an open-source tool and public leaderboard for tracking AI coding assistant token usage. It is operated by Junho Yeo. The full source, including everything described here, is available at github.com/junhoyeo/tokscale.

What Tokscale never collects

The Tokscale CLI reads your local AI assistant session files to count tokens. It does not transmit their contents. Specifically, Tokscale never collects:

  • Prompts, responses, or any conversation content
  • Source code, file contents, or file names
  • API keys or AI provider credentials

Only the aggregate numbers described below leave your machine. You can inspect exactly what would be sent before submitting anything.

Information from your GitHub account

Signing in uses GitHub OAuth with the read:user and user:email scopes. Tokscale stores your GitHub numeric ID, username, display name, avatar URL, and email address. Tokscale never requests access to your repositories.

Usage data you submit

When you run a submission, the CLI sends:

  • Token counts, broken down into input, output, cache read, cache write, and reasoning tokens
  • Estimated cost, derived from public provider pricing
  • The date range covered, and a per-day breakdown of the above
  • Identifiers of the AI clients used (for example Claude Code, Cursor, Codex) and of the models used
  • Message counts — how many messages, never their content
  • Session timing metrics: total active time, longest continuous session, maximum concurrent sessions, and session count
  • The names of MCP servers configured in your clients
  • The version of the Tokscale CLI used

Each machine you submit from is recorded as a device, identified by a device key and the device label reported by the CLI.

What is public

Tokscale is a public leaderboard, so once you submit, your profile page and leaderboard entry are visible to anyone, including people who are not signed in. That covers your username, avatar, display name, total tokens and cost, contribution graph, per-client and per-model breakdown, your devices, and any groups you join.

These profile pages are also listed in the Tokscale sitemap, which means search engines may crawl and index them. If you do not want that, delete your submitted data or your account using the options below.

Cookies

  • Session cookie — keeps you signed in after GitHub authentication
  • OAuth state cookie — short-lived, protects the sign-in flow against cross-site request forgery
  • leaderboard-sort-by — remembers whether you sort the leaderboard by tokens or cost

Analytics and service providers

Tokscale relies on these third parties:

  • GitHub — authentication
  • Vercel — hosting, and Vercel Analytics for aggregate page view statistics
  • Neon — the PostgreSQL database where the data above is stored

Advertising

Tokscale does not currently display advertising and sets no advertising cookies. If advertising is introduced, this policy will be updated before any ad code is served, and this section will name the ad networks involved. Third-party ad vendors commonly use cookies to serve ads based on prior visits to a site.

Deleting your data

Both options are available from your settings page:

  • Delete submitted data — permanently removes your submissions and device records while keeping your account
  • Delete account — permanently removes your account and every record associated with it, including submissions, devices, API tokens, and group memberships

Both are immediate and cannot be undone. You can additionally revoke Tokscale's access from your GitHub account settings at any time.

Changes to this policy

Material changes will be reflected in the “last updated” date at the top of this page. Because Tokscale is open source, the full history of this document is visible in the repository.

Contact

Questions about this policy can go to i@junho.io or the contact page.

Privacy Policy | Tokscale